intonational phonology in colloquial singaporean english*
TRANSCRIPT
Intonational Phonology in Colloquial Singaporean English*
Kelly Banciella Smemo
A thesis submitted in partial fulfillment of the requirements for the degree of Bachelor of Arts in Linguistics
Swarthmore College December 2015
Abstract
One of the primary goals of intonational phonology is to explore how meaning is assigned to an utterance through various suprasegmental features at the word level and above. This can be modeled using Autosegmental Metrical theory which creates a hierarchical prosodic structure for an utterance. How these different levels interact with each other is dependent on the language in question (Ladd, 2008).
This thesis explores Colloquial Singaporean English (CSE), a language native to the city-state of Singapore. Standard English along with Mandarin Chinese, Malay and Tamil are recognized as the official languages of the area. This puts CSE in a stigmatized role, despite its use in almost all spheres of life, barring government policy and writing in general. The dialect differs greatly from standard forms with heavy borrowing from languages such as Mandarin, Hokkien and Tamil on phonetic, syntactic and semantic levels (Harada, 2009).
Within CSE, there is contention on how to accurately describe its intonational system due to the lack of empirical data. One model, developed by Ng posits that each syllable of the word is assigned a specific tone based on it's place in the word (e.g. Ng, 2011). Another model, argued for by Chong, uses Autosegmental Metrical theory to break up utterances into accentual phrases (e.g. Chong, 2013).
In order to test the merit of each model, I collected novel data from two consultants focusing on target words of varying syllable length. I then com pared the two models using this data. Ng's Model failed to accurately portray most of the data collected. Chong's Model was more successful but in need of modification. I proposed two modifications, one that would treat prefixes as their own prosodic unit (in line with the existing findings on the prosodic behavior of prefixes) and another that allowed for multiple prosodic units that determine intonational contours within a multisyllabic word greater than four syllables.
*1 would like to thank my advisor, Prof. Byron Ahn for helping me through every step of the thesis process. I would also like to thank Prof. Ted Fernald for being my second faculty reader. In addition, I would like to thank Prof. Nathan Sanders for providing aid in the initial steps of this project. Lastly, this thesis would not have been possible without the contribution of my two consultants, PK and RT.
1
Contents
1 Introduction
2 Background 2.1 Intonational Phonology. 2.2 Colloquial Singaporean English 2.3 Current Competing Models
3 Data 3.1 Consultant Description 3.2 Data Overview ... . 3.3 Data Description .. .
3.3.1 One and Two Syllable Words 3.3.2 Three Syllable Words ..... 3.3.3 Four or More Syllable Words.
4 Analysis 4.1 Prefixes as Unique Prosodic Units ...... . 4.2 Multisyllabic Words and the Accentual Phrase 4.3 Ternarity.. 4.4 Exceptions.
5 Conclusion
6 Appendix
References
2
3
3 4 9
11
16 16 17 19 19 24 26
28 29 31 34 36
37
41
67
1 Introduction
Colloquial Singaporean English (CSE) is an important language to study due to its contact
with many other languages, which has influenced it greatly. Because of this, linguists have
examined the language to discover how it differs from Singaporean English in its phonology,
syntax and semantics. These subjects have been studied at length (Harada, 2009; Leimgru
ber, 2011; Tay, 1993). However, within the field of phonology, one continuing mystery of
CSE is its intonation system.
This glaring gap in the field is due, in part, to the general lack of data on the subject.
In this paper, I enter the current conversation and provide the subfield some of the data it
needs. In addition, I evaluate two current models on the intonational phonology of Colloquial
Singaporean English, Chong and Ng's Models, and compare them. Ng's Model is lexically
tone based while Chong's Model relies on AM Theory. In comparing them, I found Chong's
Model to be a better predictor of the majority of the data. However, it does need a few
modifications in order to be fully representative of the data. Because of this, I propose two
important modifications to one of the models.
The two modifications proposed call for a re-evaluation of what accentual phrases consist
of. Chong's Model initially has APs consisting of an entire content word and its related
function words. In addition to this I propose that prefixes should be considered their own
APs as well. Furthermore, longer multisyllabic words have the ability to break up into
multiple APs, with a preference toward ternarity, or groupings of three.
2 Background
3
In order to fully comprehend how intonation is used in Colloquial Singaporean English, we
need to have a fundamental understanding of the field of intonational phonology and some
background on Singaporean English itself. In addition, it is important to understand the
current dialogue on intonation within Singaporean English. In this section, I briefly explain
intonational phonology as it pertains to this paper, touch upon Singaporean English and
introduce the two current competing models for intonation within Colloquial Singaporean
English.
2.1 Intonational Phonology
Intonation "refers to the use of suprasegmental phonetic features to convey 'postlexical' or
sentence-level pragmatic meaning in a linguistically structured way." (Ladd, 2008). In order
to fully understand this definition we must understand two key terms within it: supraseg
mental and postlexical.
For the purposes of this paper, the term suprasegmental will be defined as "the features
of fundamental frequency (FO), intensity and duration" (Ladd, 2008). In other words, we
will be looking at the pitch track of the utterance, noting where it rises or falls intensely and
for how long. The fact that others have defined this term differently will not be discussed
within this paper.
The term postlexical is used to define the idea of meaning existing above word level.
Lexical items create meaning at word level, while postlexical units contribute meaning above
this level. In this thesis we will be focusing on postlexical phonological items. Intonation is
used to add meaning to phrases or entire utterances. For example, the difference between
the declarative sentence I was chosen for the constitutional convention. and the question I
4
was chosen for the constitutional convention? is intonation. In the latter, the speaker rises
in tone in order to form the question (Gunlogson, 2008). Intonational features are never used
to describe lexical differences in languages. However, some languages may use the phonetic
basis of intonation, such as stress or tone, to differentiate lexical items. It is important
to draw a distinction between intonational phonology and the use of phonetic intonational
items for differentiating lexical items.
It is also important to note that intonation is linguistically significant. This can be
demonstrated for in English, where pitch-accent can change the fundamental meaning of a
sentence but not the fundamental meaning of a word. For example, Figure 1 and Figure 21
both show the same sentence, only differing in intonation. In Figure 1, the pitch-accent is
on the word steal. If we look at the figure, we see a dramatic rise in pitch track2 on the
word steal. The blank spot at the beginning of the word is due to the voiceless fricative Is],
which has no pitch. Other gaps in the pitch track may be due to Praat3 having difficulty
reading the item. In addition to the dramatic rise in pitch, the duration of the word is longer
than in Figure 2. Thus, we know the pitch accent, which is only associated with stressed
(and thereby longer) words, is on the word steal. Figure 2 has a pitch accent on the word
money. We see this on the dramatic rise in the first syllable of the word. The first syllable
of the word money is the lexically stressed syllable, and pitch-accents in English typically
find themselves on the stressed syllable of the word. In addition, the word itself is said for
a longer amount of time. Because of this difference in intonation between the two example
sentences, the meaning changes. In the first, the speaker is implying that they did not steal
1 Both sentences in Figures 1 and 2 are spoken by me. 2The pitch track is the FO line that marks intonation on an utterance 3Praat is the program being used to analyze the data.
5
the money, but rather, did something else to the money. In Figure 2, the speaker is implying
that they didn't steal the money, but rather, they stole some other item. The meaning of
the words themselves have not changed, but rather the meaning of the entire utterance.
151117-214323 151117-214323 553284555 8 37906288
30 0
250 25 0
200 1-
~ 150
0 i~ ~ w~ ~J 0 !
20
g 15
jj
" 100 10 0
50+-+---+-------+---+------1 0
I chdn't steal money I chdn't steal ili. money
3809 5 536 6 998 8379 T,me (sl T,me (sl
Figure 1: Pitch-accent on steal Figure 2: Pitch-accent on money
It is important to note that pitch-accent should not be equated with stress, as it is easy
to confuse the two within English. The source of this confusion can be found in the fact
that, in stress-accent languages (like English), pitch-accent is found on the stressed syllable
of the prominent word. Pitch-accents can only be found where there is stress, but where
stress is found is determined independently of this. Pitch-accent is found phonologically on a
higher prosodic level while lexical stress is found word level. Lastly, pitch-accents are found
phonetically and phonologically in languages that do not have stress (Ladd, 2008).
These pitch events can be the result of different types of intonation systems including:
stress-accent, lexical tone and pitch-accent. English is just one example of a stress-accent
language. In these languages, stress is used to differentiate the meanings of lexical items. For
example, the verb permit and the noun permit are only distinguishable through their stress
pattern. The stress is on the second syllable of the verb form and on the first syllable of the
6
noun form (Ladd, 2008). Pitch-accent is used to put emphasis on the prominent words, and
can only be found on the stressed syllable of the word, as noted previously. This is why, in
Figure 2, the pitch accent is only found on the first syllable of the word money.
Lexical tone languages use tone to differentiate lexical meaning between words. One
such language is Mandarin Chinese. For example, the words hua 'flower' and hua 'speech,
language' are identical in every way except for their tone. The first has a high level pitch
while the latter has a falling pitch (Ladd, 2008). In this way, tone is heavily tied with the
lexicon of the language rather than the postlexical intonation system.
Finally, pitch-accent languages, such as Japanese, are usually defined as employing an
intonation system which "uses pitch to mark certain syllables in the speech stream" (Venditti,
2005). In other words, pitch-accents are a lexical part of the word, like tone is in Mandarin
Chinese. In addition, unlike Mandarin, but similarly to English, most syllables are not
associated with a pitch event. Unlike stress-accent languages, pitch-accent does not have
any prominence-lending attributes. It is also important to note that pitch-accents do not
occur on every word. For example Figures 3 and 4 show identical phrases differentiating
only in pitch accent. Figure 3 is not accented while Figure 4 has a pitch accent on ue. We
can see this clearly by the dramatic rise in the pitch track on that specific syllable. This
difference in pitch accent creates two very different lexical meanings, with the first meaning
'something to plant' and the latter meaning 'the ones who are starved'. (Venditti, 2005:2)
It is important to differentiate these different language types in order to understand
which category Colloquial Singaporean English belongs in. One thought could be that it is
a stress-accent language, as it is a variant of English. Another thought could be that it is
a lexical tone language, especially since it has extensive contact with and borrowing from
7
u e u m 0 n 0
Figure 3: unaccented uerumono 'something to plant'
-• -. • • • • '10 • • .. • - '10
......... -.. u e' r u m 0 n 0
Figure 4: accented ue)rumono 'the ones who are
starved'
Mandarin Chinese. Lastly, there is a possibility that Colloquial Singaporean English is a
pitch-accent similar to Japanese because of it's use of pitch-accents in its intonation system.
This topic will be further discussed in Section 5.
We can model all of these different intonational systems by using Autosegmental Metrical
(AM) theory. This theory proposes two major points. One, that intonation manifests itself
in local events and transitions. In other words, one does not have to describe each minute
transition on every syllable in intonation, but rather the key events that make it up which are
distributed more sparsely in the utterance. Secondly) it argues that utterances are comprised
of a hierarchichal prosodic structure made up of different levels. These levels include the
intonational phrase (IP), the intermediate phrase (ip) and the accentual phrase (AP). In
addition) there is a word level and syllable level4 . We can see how these levels are structured
in Figure 5.
If we read Figure 5 from bottom to top we begin with the smallest unit) the syllable.
4Which levels are present is dependent on the language in question. In addition, there may be additional prosodic levels outside the ones discussed here but those are irrelevant to this thesis.
8
Syllables are marked with a (S) if they are stressed and a (s) if they are unstressed. Syllables
build up to words. Words then make up accentual phrases (AP). An AP is made up of a single
content word (Wc) and its accompanying function words (Wf) if applicable. For example,
the term the monkey would be one AP. Above the accentual phrase is the intermediate
phrase (ip) which can have one or more APs within it. Finally, the intonational phrase (IP)
encompasses the entire utterance.
IP
~ lp (ip)
~ AP (AP)
~ (Wf) Wc
~ ... S s s ...
Figure 5: AM Theory Hierarchical Structure
How intonational targets associated with these different hierarchies affect or override each
other depends on the language (Ladd, 2008). We will examine in depth one possible AM
model for Colloquial Singaporean English in Section 2.3.
With this brief background on intonational phonology, we are now able to use these
models to collect and analyze data from speakers.
2.2 Colloquial Singaporean English
Singaporean English (SgEng) is a variety of English spoken in Singapore, a city-state south
of Malaysia (Harada, 2009). Although tiny, this city is ethnically and linguistically diverse.
It has four official languages: Singaporean English, Mandarin Chinese, Malay and Tamil.
Of these four, SgEng has the most prestige. It is used primarily in formal situations and all
9
written documents. However, the focus of this paper is not on SgEng but rather on its more
colloquial variant, Colloquial Singaporean English (CSE).
CSE is spoken in complementary distribution to SgEng. It varies significantly from SgEng
in its syntax and phonetics, due to the influences of Mandarin, Hokkien (another Chinese
dialect), Malay and Tamil (Harada, 2009). These variations include but are not limited to
copula deletion, article deletion and its lack of verb marking (Harada, 2009). Because of
these variations, the language has been stigmatized by the government, which has pushed
for its eradication; however, many of the younger generation uphold the collquial variant as
an important part of their cultural identity (Harada, 2009).
One way CSE differs from SgEng is through article deletion. We can see in example (la)
a sentence in Singaporean English. We can compare this to example (lb) with the CSE
equivalent. In (lb) the article a is not pronounced with the noun phrase, while in (la) it
must be there to be grammatical. (Data (1-3) taken from Tay, 1993:27-35)
(la) May I apply for a car license? (lb) May I apply for car license?
(SgEng) (CSE)
CSE also differs from SgEng through its verb marking. CSE tends to not inflect verbs
for tense, and instead, relies on syntactic structures and context to derive meaning. For
example, in (2a) we see a sentence in SgEng that marks the verb go with a third person
singular marker. However, in the CSE equivalent (2b), go is left in the infinitive, unmarked
for tense.
(2a) He always goes there every Saturday. (2b) He always go there every Saturday.
(SgEng) (CSE)
In addition, CSE has multiple instances of "copula deletion", differentiating it from the
standard dialect. In sentence (3a) we see the SgEng sentence, which includes the copula
10
Particle ah hah hor lah leh lor mah what/wot meh ya
Meaning tentative marker, continuation marker question marker attempts to garner support for a proposition mood marker, appeals for accommodation marks a tentative suggestion/request indicates obviousness or resignation marks information as obvious marks obviousness and contradiction indicates scepticism conveys (weak) emphasis and uncontroversiality
Table 1: Selected Particles of CSE
to be in order to be grammatical. In the CSE version in example (3b), this verb IS not
pronounced and the sentence remains grammatical.
(3a) (3b)
My handwriting is not clear. My handwriting not clear.
(SgEng) (CSE)
Furthermore, CSE varies with its vocabulary and heavy borrowing from other nearby
languages. For example, CSE uses a number of clause final discourse particles, most likely
borrowed from Hokkien or Cantonese, whose exact meanings are still being disputed (Leim-
gruber,2011). In Table 1 we can see a number of these particles as well as their corresponding
meaning according to Leimgruber. Table 1 has been reproduced exactly from Leimgruber
(Leimgruber, 2011:9).
The intonational phonology of Colloquial Singaporean English has not been studied to
the extent as many other parts of the language. The purpose of this paper is to explore this
topic further.
2.3 Current Competing Models
The current dialogue on Colloquial Singaporean English's intonation system is limited. In
part due to the relatively small number of research publications on the topic, there IS m-
11
sufficient data to make all the necessary observations in order to establish a more complete
model of the language's intonation. There are two com peting models that attem pt to explain
CSE's intonation system, one is based on the influences of Mandarin Chinese lexical tone,
and the other on AM Theory.
The former has been developed in several works by E-Ching Ng (Ng, 2011; Ng, 2012).
Ng's Model is based on the idea that CSE is a lexical tone language, like Mandarin Chinese.
The model states that every syllable of a word carries a tone based on its placement within
the word.
Ng presents her tone assignment generalizations as follows (Ng, 2012)5:
• H is assigned to the final syllable of the prosodic word. • L is assigned to initial unstressed syllables. • M is assigned to all remaining syllables.
She also includes examples of tone assignment with multiple words of varying syllable
length. This table is reproduced in Table 26 (Ng, 2012:87). Examples (a)-(e) demonstrate
words with initial stressed syllables and therefore no low tones, while examples (f)-(j) show
words with initial unstressed syllables. Although Ng provides this table, she provides no
pitch tracks for words longer than three syllables, nor does she explicitly discuss them.
In addition, these multisyllabic examples are only measured after they are produced in
isolation rather than in a sentence. This is problematic because words spoken in isolation
may behave differently than when spoken in a full utterance. As discussed in Section 2.1,
standard assumptions of intonational phonology include that there are multiple hierarchies
encoding meaning into a phrase and these hierarchies may affect each other. If a word is
5 Where H stands for a high tone, L for a low tone and M for a mid tone. 6It is important to note that this table has been reproduced exactly from Ng's paper, including the ways
in which stress is marked within it.
12
a. ~e 'H f ma'chine L'H b. money 'MH g. hi 'biscus L'MH c. 'elephant 'MMH h. A'merica L'MMH d. 'Indo'nesia 'MM'MH , elec'tronics LL'MH e. 'minimi'sation 'MMM'MH ] res'ponsi 'bility L'MM'MMH
Table 2: Ng's Tone Assignment Data
spoken in isolation, it is spoken utterance finally, and may be affected by multiple hierarchie.s
including the IF, ip and AP. On the other hand, if the word is spoken in the middle of an
utterance, it will not be affected by the IF or even the ip.
Ng's model is ba.sed on a rather limited set of CSE data, One of the few presentations of
a pitch track of a sentence is reproduced in Figure 6 (Ng, 2011:36). Each syllable is marked
by a tone according t o the generalizations discussocl previously. The word cannot1 starts
with a low t one on the first syllable and ends with a high tone on the last syllable. The word
minimum has the first two syllables notated as mid tones and ending with the last syllable
as a high tone. Thus, we clearly see a low to high or mid to high pattern on each word. This
is predicted accurately with her model.
F igure 6: Example Sentence Transcribed by Ng
Although Ng never mentions function words in her analysis, she marks the function words
in her examples as having a low tone. We soo this on the word from in Figure 6. I will follow
suit when I test her model in Section 3.3.
tIt is interesting to note that cannot is treated as one prooodic word. It would be interes ting to ilNestigate if this were true for other words such as will not. Unfortunately, this topic is outside the scope of this thesis.
13
One problematic aspect of Ng's analysis is her use of the term lexical tone. The phe
nomenon she describes within her paper, of syllables having an assigned tone based on
location, does not mirror what one would find within a lexical tone language (in fact, this
sounds more like a pitch-accent language). In particular, there does not seem to be any min
imal pairs of CSE that differ only by tone. This makes it highly unlikely that the language
is a lexical tone language.
In contrast to Ng's lexical tone model, Chong uses AM theory to devise a model (Chong,
2012; Chong, 2013; Chong & German, 2015). He proposes that CSE's intonation pattern is
the result of tone targets associated with higher prosodic levels than the word level. Although
some of the work is co-authored, I will be referring to this model as Chong's Model.
Chong's Model proposes that the low to high pattern Ng has observed is not from the
lexical properties of the word and its syllables but rather from the Accentual Phrase (AP).
As mentioned in Section 2.1, the AP consists of either one word, or a word and its associated
function words. In this way, he specifically accounts for function words within his model,
unlike Ng. Chong's Model is reproduced in Figure 7 (Chong, 2013:2)
Chong also provides for more levels above the AP, including the Intermediate Phrase (ip)
and the Intonational Phrase (IP). The boundaries for the AP are defined with the following
notation: an initial low tone (aL) and an ending high tone (Ha). Both the ip and IP are edge
marked by either a low or high tone. They can be differentiated in the notation with either
a dash (-) or percentage mark (%) respectively. For example, a low IP-edge tone would be
notated as L% while a low ip-edge tone would be marked as L-. (Chong, 2013)
Chong's Model also differs in that it allows for a low tone pitch accent (L*) that can be
found within an AP. The inclusion of a low tone pitch accent (L*) creates a key difference
14
aL
IP
lp (ip)
AP (AP
Wc
~ ... S s s ...
I
(L*) Ha L- or H-
Figure 7: Chong's Model
L% or H%
in the two models. For one, the L* can only occur on a stressed syllable. In Ng's model, it
is impossible for a stressed syllable to have a low tone. However, it will become apparent in
the data in Section 3.3 that words with a stressed low tone do, in fact, exist. An example of
one can be taken from Chong's own paper, reproduced in Figure 8 (Chong, 2013:3).
In Figure 8 we see that the utterance is comprised of three APs framed by an aL and Ha
notation. Each has a low pitch accent (L*) on a syllable of the word. The utterance ends
with a low ip (L-) and a low IP (L%). In this way we see how both word boundary and edge
boundary (ip and IP) can affect the pitch of an utterance.
Because of the relatively low amount of research in this area, Chong's analysis (like Ng's)
is built upon a small empirical base, and it must be tested on a larger data set. One major
gap in all previous work is systematic investigation of multisyllabic words greater than three
syllables. Ng touches upon the issue very briefly in an example, but presents the words in
isolation rather than in a sentence (Ng, 2012). This is a gap I address with an experimental
design that manipulates syllable length as a variable.
15
L· , .. H. IrLI<
-Figure 8 &i:,.mple Sentence Tronscribed l:y Chong
3 Data
3.1 Consultant Description
The d,.b used in t his rese",ch wo.:; collected from tw o consulb nls They will be referred to
o.s PK ,.nd FIT in this p"per To prElServe onorvmity, these inili<ik were formed by combining
,. rondom poir of lellers Both consul\onls ne n,.t ive Sing,.po reon English speuers
PK is fluent in Mond"'in Chinese, is ,.dequoJe in M,.loy ond Hokkien, ,.nd ho.s some
knowledge of hponese ond ContonElSe She is 22 ye",s old, born ond r.:.;ed in Penon~
M,.loysi,. However, she spent her Middle ,.nd High school yen" in bo",ding school in
Sing,.pore She is ethnicolly Str,.ils-born ChinElSe
FIT wo.s born in Sing,.pore ond spent his life until university in the city sbte He ho.s on
adequate grasp of Chinese and some knowledge of Malay, Indonesian and Cantonese. He is
23 years old and ethnically Chinese.
3.2 Data Overview
Each consultant was given a list of twenty SIX sentences to read aloud. Each sentence
comprised of a carrier phrase with a target word, also known as a variable word, which
varied in syllable length. A carrier phrase is a sentence used to introduce a target word.
The phrase remains consistent throughout the data set while the target word is changed to
test a certain variable (in this case, syllable number). Target words were chosen based on
sonority, because a more sonorous word would show pitch fluctuations easier. The majority
of the sentences had the target word in the middle of the utterance, so as not confuse the
word boundary with the edge boundary. However, a few sentences purposefully placed the
target word at the end of the sentence in order to see if there was a difference.
Sentences (1), (5), (10) and (15), are examples from the complete set of sentences found
in the Appendix, used here to illustrate the concept of carrier phrases and target words. The
phrase Say {insert] again is the carrier phrase for these five sentences. The target word is
placed within this phrase. The target words within this set are all phonetically similar and
primarily differ in syllable length. In sentence (1) the target word Min is made up of one
syllable, in sentence (5), mini is made up of two syllables, in sentence (10) minimal is three
syllables and, finally, sentence (15) has the four syllable word minimally. The meanings of
the target words do not matter.
(1) Say Min again. (5) Say mini again. (10) Say minimal again. (15) Say minimally again.
17
In Section 3.3 each sentence is labeled with both Ng and Chong's Models. For Ng's
Model, each syllable is marked for tone based on its placement in the syllable, as discussed
in Section 2.3. Similarly, Chong's Model is notated using his notation of L* for pitch accents
within APs, aL and Ha to bracket APs, either L- or H- to mark the ends of ips if needed
and L% and H% to mark the ends of IPs.
The data in Section 3.3 is presented in a way that displays the FO height for each syllable
or prosodic unit in the sentence elicited by both consultants. The models are presented
side by side for ease of comparison. The values of tone height were taken from reading
spectrograms. Usually, this was done by paying attention to the pitch track generated by
Praat. However, due to the background noise present in most of the recordings, this pitch
track was not always a reliable indicator of tone height. In these cases, I impressionistically
evaluated the sound files to identify pitch at different times. This was made easier in the
stark low to high patterns the speakers regularly had in their speech.
An example of one pitch track is found in Figure 9. This is the pitch track for sentence
(7) spoken by consultant RT8 The dark lines show the pitch track (FO line) of the utterance.
The first tier displays Chong's model, the second tier displays N g's Model and the bottom tier
has a gloss of the sentence. Chong's Model marks the boundaries of phrases. For example,
we can see the boundaries for the accentual phrase the word bordered by aL and Ha. The
entire utterance is then marked by the edge boundaries for the ip and IP which are both
low in this example. Ng's Model splits the utterance into syllables and then marks each
syllable with a tone. For example, the is marked with a low tone while word is marked with
a high tone. In this pitch track we can clearly see that both models are able to capture the
8 A full listing of pitch tracks for each sentence can be found in the Appendix.
18
repeating rising patterns on each word. This is especially apparent in the word and believe.
Both words begin low and end high in the pitch track.
1.35244717 190
, • • • 150 • ..... •
/' • • •
I vv • , ~ " :. ! .",;......; :\
~ i i \ ' . • " 100
75 I I I I I I
aL HaaL Ha aL HaaL IThb% I I I I I I I I I
H L H L H L H
say the word believe agam
0.09771 .382 Time (s)
Figure 9: Say believe again spoken by consultant RT
3.3 Data Description
The data in this section has been divided based on syllables. For each section we will compare
Chong and Ng's models. We will point out contradictions and any problems with how each
model handles the data. However, we will not go into depth on these issues until Section 4.
3.3.1 One and Two Syllable Words
For words that consist of one and two syllables, both Chong and Ng's models are accurate,
with varying levels of clarity. One thing of note, however, is that Chong explicitly accounts
for function words by placing them within APs. Ng, on the other hand, does not mention
them. However, in her own examples she notes them with a low tone, as I have replicated
19
below.
In order to understand how both models are able to account for one syllable words, we
should examine the data in depth. Figure 10 shows the pitch track for sentence (1) Say
Min again, said by consultant RT. The first tier of the figure displays Chong's Model using
AM theory, the second tier displays Ng's Model, and the third tier displays a gloss of the
sentence.
In Figure lOwe can clearly see the low to high pattern in both the target word Min and
in the word again in the pitch track itself. This pattern is best expressed with Chong's Model
which notes that the beginning of the AP is low (aL) and ends in a high (Ha). This differs
from Ng's model which simply states if the syllable is high (H) or low (L). In one syllable
words, such as Min, this rising pattern is not easily communicated. It is only notated as
high tone, making the comparison with multisyllabic words a harder one to make.
We see this same pattern repeated in all of the one syllable word sentences, for both
consultants PK and RT. The results for those sentences are reproduced below in table form.
One thing to note is the phrase the word and how it behaves. In Chong's Model, it makes up
one AP rather than two, as the is a function word associated with word. In Ng's Model, the
representation is still accurate as we have chosen to mark function words with a low tone.
The data (la)-( 4b) list all of the sentences with one syllable target words. The data shows
that both speakers produced the same intonational patterns. In addition, both models were
able to predict these patterns. (la)-(4a) apply Ng's Model's predictions to the data while
(lb )-( 4b) use Chong's Model's predictions. The Appendix has a listing of all pitch tracks
for each sentence and consultant, annotated with both models.
20
106643871 210
: .:
I • : :
150 • ~ .... I . ·-N J ~~ •.
g 100
I I I I I I ~ p, 75
aL Ha aL Ha aL Ha HE=I%i
I I I I I I I I
H H L H
Say Min agam
0.04197 1075 Time (s)
Figure 10: Say Min again spoken by consultant RT
(la) Ng Model (lb) Chong Model Say Min agam. Say Min agam.
PK H H LH aL Ha aL Ha aL Ha H-H% RT H H LH aH La aL Ha aL Ha H-H%
(2a) Ng Model (2b) Chong Model Say live agam. Say live agam.
PK H H LH aL Ha aL Ha aL Ha H-H% RT H H LH aL Ha aL Ha aL Ha L-L%
(3a) Ng Model (3b) Chong Model Say the word bee agam. Say the word bee agam.
PK H L H H LH aL Ha aL Ha aL Ha aL Ha L-L% RT H L H H LH aH La aL Ha aL Ha aL Ha L-L%
( 4a) Ng Model ( 4b) Chong Model Say the word may. Say the word may.
PK H L H H aL Ha aL Ha aL Ha L-L% RT H L H H aL Ha aL Ha aL Ha L-L%
Similar to one syllable words, two syllable words are also well described by both models.
The previous issue of Ng's where the single tone syllable did not accurately portray the rising
21
pattern of the word is no longer an issue. We can see this in Figure 11, which shows the
pitch track for sentence (7) Say the word believe again spoken by consultant RT.
Again, the low to high pattern is easily apparent. In the target word believe, we start
with a low tone and end in a high. It is important to note that the pitch track is poorly
drawn for the last word of the sentence again. This is likely due to the combined problem
of the loud background noise as well as the fact people tend to go creaky (which disrupts
the pitch) at the ends of utterances (Ladd, 2008). This makes it difficult for Praat to track
the pitch in the acoustic signal. However, when listening to the audio recordings and using
impressionistic annotation, the low to high rise is easily discernible. This is still discernible
even in the last word of the sentence with a possible subsequent fall due to a L- or L%
boundary tone.
1.35244717 190
, • • • 150 •
" •
/' • • •
I
~ \
"
""'" ! .".;.....: .'\ :.
~ ! ! \ '. • '" g 100
I I I I I I ~ p, 75
aL HaaL Ha aL HaaL IThb% I I I I I I I I I
H L H L H L H
say the word believe agam
0.09771 .382 Time (s)
Figure 11: Say the word believe again spoken by consultant RT
22
Sentences (5a)-(8b) include all instances in which the target word has two syllables. The
patterns we have laid out previously repeat themselves. Each word has a low to high rise,
with the exception of the function word the. However, this is accounted for in Chong's Model
by including it in the AP the word. In Ng's model, we have accounted for this by marking
it with a low tone.
Unlike in sentences with one syllable, Ng's model now accurately portrays the rising
action of the intonation of these sentences. Because they have two syllables, we are able
to mark the initial syllable as low or mid and the final syllable high, accurately portraying
the rise. We mark the syllable as a mid tone if it is stressed, such as in the word mini and
living. Otherwise, we mark the first syllable as a low tone, such as in the word believe. This
is another aspect in which the two models differ, as Chong's Model predicts no difference
between the intonational patterns of these words. Within the data collected it is difficult to
state whether or no there is a difference in the intonation of these words.
The data (5a)-(9b) list all the sentences with two syllable target words. Again, the data
shows that both speakers produced the same intonational patterns. In addition, both models
were able to predict these patterns. (la)-( 4a) apply Ng's Model's predictions to the data
while (lb )-( 4b) use Chong's Model's predictions. One can consult the Appendix for specific
pitch tracks of these sentences.
(5a) Ng Model (5b) Chong Model Say mml agam. Say mml agam.
PK H MH LH aL Ha aL Ha aL Ha L-L% RT H MH LH aL Ha aL Ha aL Ha L-L%
(6a) Ng Model (6b) Chong Model Say living agam. Say living agam.
PK H MH LH aL Ha aL Ha aH La L-L% RT H MH LH aL Ha aL Ha aH La L-L%
23
(7a) Ng Model (7b) Chong Model Say the word believe agam. Say the word believe agam.
PK H L H LH LH aL Ha aL Ha aL Ha aL Ha L-L% RT H L H LH LH aL Ha aL Ha aL Ha aL Ha L-L%
(8a) Ng Model (8b) Chong Model Say the word maybe. Say the word maybe.
PK H L H MH aL Ha aL Ha aL Ha L-L% RT H L H MH aL Ha aL Ha aL Ha L-L%
(9a) Ng Model (9b) Chong Model Say the word Image agam. Say the word Image agam.
PK H L H MH LH aL Ha aL Ha aL Ha aL Ha L-L% RT H L H MH LH aL Ha aL Ha aL Ha aL Ha L-L%
3.3.2 Three Syllable Words
In words with three syllables, issues begin to appear in Ng's Model while Chong's Model
remains strong. Some three syllable words in CSE have a low pitch target in the middle of
the word. This is allowed for in Chong's Model with a low pitch accent (L*) but not in Ng's.
For example, in Figure 12, we see the three syllable target word believer. In Chong's Model
this is transcribed as having an initial AP low (aL) and a final AP high (Ha). However,
within the word, the second syllable has a low pitch accent, shown as L* In Ng's model, it
is impossible for a stressed syllable to have low tone. Instead, we must notate the syllable
as a mid tone, regardless of the fact the pitch track clearly shows it remains a low tone.
Therefore, this mid tone does not accurately represent the pronunciation of the word and is
a problem in Ng's analysis.
This low pitch accent variable occurs in some of the three syllable target words within the
dataset. Examples (12b) and (13b) both show that the words believer and maybelline both
have a low pitch accent (this finding is consistent between both speakers). In example (llb),
however, we see that consultant PK has a low pitch accent in the middle syllable of livable
but RT does not. Examples (lOb) and (14b) show two three syllable words, minimal and
24
2.39144676 350
0 l . .' .. .~
.'\ A \ ~ • • '1 ,. :
• ....... • • .. ,. .i
300
250 N e; ""
I I I I I . .@
200 p,
~ HaaL Ha aL L* Ha aL Ha LI, , I I I I I I I I I
H L H L M H L H
say the word believer agam
0.8014 2.418 Time (s)
Figure 12: Say the word believer again spoken by consultant PK
imagine which have no low pitch accent for both speakers. For these latter two sentences,
Ng's Model works equally as well as Chong's ModeL There is a debate surrounding the
existence of the low pitch accent (L*). Chong states that it is optional and that it seems
to only occur in the prominent word of the sentence, although further research is needed to
fully understand it (Chong, 2012).
The rest of the data within these sentences remains consistent with two syllable and one
syllable sentences. The words say and again both have an initial low and final high. The
phrase the word continues to work as an Ap 9
(lOa) Ng Model (lOb) Chong Model Say minimal agam. Say minimal agalil.
PK H MMH LH aL Ha aL Ha aL Ha L-L% RT H MMH LH aL Ha aL Ha aL Ha L-L%
gIn sentence (11), RT's data is difficult to discern. Because of this, the missing values are marked with anX
25
(lla) Ng Model (llb) Chong Model Say livable agam. Say livable agam.
PK H MMH LH aL Ha aL L* Ha aL Ha L-L% RT H MMH LX aL Ha aL Ha aL X XX
(12a) Ng Model (12b) Chong Model Say the word believer again. Say the word believer again.
PK H L H LMH LH aL Ha aL Ha aL L* Ha aL Ha L-L% RT H L H LMH LH aL Ha aL Ha aL L* Ha aL Ha L-L%
(13a) Ng Model (13b) Chong Model Say the word maybelline. Say the word maybelline.
PK H L H MMH aL Ha aL Ha aL L* Ha H-H% RT H L H MMH aL Ha aL Ha aL L* Ha L-L%
(14a) Ng Model (14b) Chong Model Say the word Imagme agam. Say the word Imagme agam.
PK H L H LMH LH aL Ha aL Ha aL Ha aL Ha L-L% RT H L H LMH LH aH Ha aL Ha aL Ha aL Ha L-L%
3.3.3 Four or More Syllable Words
There are only two instances in which the low pitch accent is present in a four syllable word.
In example (15b), consultant PK pronounces minimally with a low pitch accent on the third
syllable but RT does not. Conversely, in sentence (17b) consultant RT pronounces imagining
with a low pitch accent on the third syllable while PK does not. In all other sentences with
four syllable target words, (16b)'s believable, (18b)'s malaria and (19b)'s Romania, neither
speaker pronounces the word with a low pitch accent.
(15a) Ng Model (15b) Chong Model Say minimally agam. Say minimally agam.
PK H LMMH LH aL Ha aL L* Ha aL Ha L-L% RT H LMMH LH aL Ha aL Ha aL Ha L-L%
(16a) Ng Model (16b) Chong Model Say the word believable again. Say the word believable again.
PK H L H LMMH LH aL Ha aL Ha aL Ha aL Ha L-L% RT H L H LMMH LH aL Ha aL Ha aL Ha aL Ha L-L%
(17a) Ng Model (17b) Chong Model Say the word imagining again. Say the word imagining again.
PK H L H LMMH LH aL Ha aL Ha aL Ha aL Ha L-L% RT H L H LMMH LH aL Ha aL Ha aL L* Ha aL Ha L-L%
26
(18a) Ng Model (18b) Chong Model Say malaria agam. Say malaria agam.
PK H LM MH LH aL Ha aL Ha aL Ha L-L% RT H LM MH LH aL Ha aL Ha aL Ha L-L%
(19a) Ng Model (19b) Chong Model Say Romania agam. Say Romania agam.
PK H LM MH LH aL Ha aL Ha aL Ha L-L% RT H LM MH LH aL Ha aL Ha L- aL Ha L-L%
In multisyllabic words greater than four, more problems begin to emerge for both Chong's
and Ng's Models. In Figure 13 we se the five syllable target word livability spoken by
consultant RT. Here there is a low pitch accent found on the second syllable. This is not
predicted accurately in either model. In Chong's model, the L* can only be placed on a
stressed syllable, and the second syllable of livability is unstressed. In Ng's model, a low
pitch accent within a word is not possible. In both cases, this gives the false representation
that the entire word is a continuous rising pattern. On the other hand, other five syllable
words behave as would be predicted by both models, such as the word radiology in (25).
This same pattern seems to repeat in other six syllable words. For example, in Figure
14 we see consultant PK's pronunciation of memorabilia. Ng's Model does not accurately
predict the sudden drop in tone in the middle of the word. Chong's Model could make use
of the low pitch accent, however, it does not accurately predict the pronunciation. The first
three syllables of the word have an initial low and then rise to a high. The fourth syllable
has a low pitch target and the rest of the word displays the rising action once more. This
initial rise during memora-, before the L*, is not accurately portrayed in Chong's Model.
In addition, the majority of the sentences with six syllable words contain this pattern.
For the words, livability (20), unimaginable (21), developmentally (23),memorabilia (24)
and biodegradable (26), both speakers have an initial rise in the word, followed by a low
27
1.34182703 200
!~ . .
-~~., -"'" . ~ ~i , \~. .. : .. ~ :
150
g 100
75 I I I I I I
L HaaL Ha aL Hdb% I I I I I I I
H L M M M H L H
say livability agam
0.04744 1.352 Time (s)
Figure 13: Say livability again spoken by consultant RT
pitch target within the same word, whcih seems to reset the low to high rise. In addition,
consultant PK has this in the word developmentally in example (22).
Because of this consistent issue in both models for multisyllabic words greater than four
syllables, it is necessary to revise one of the models in order to account for this. This will
be discussed in Section 4. In addition, it is difficult to show the errors in the prediction in
table format. As such, refer to the appendix to see the pitch tracks for sentences (20)-(26).
4 Analysis
It has become apparent by looking at the data, that both Chong and Ng's models are not
entirely adequate for describing CSE's intonational system. For one, Ng's model only works
for two syllable words and greater if the word does not contain a low pitch accent. Because
it is problematic for the majority of the data with these longer words, I will be setting it
28
2.10206726 350
,;- i~ :.,j
I 300
• •
\.. ~ ...;;: -.. .. , '" . ,
~ f
\
250
200 I I I
aL HmL L* HmL HL L , I I I I I I
H L M M M M H L H
say memorabilia agam
0.332 2.102 Time (s)
Figure 14: Say memorabilia again spoken by consultant PK
aside. Chong's model on the other hand, is able to describe almost all of the data accurately,
including the examples where Ng's Model fails. However, Chong's model contains gaps in
respect to multisyllabic words greater than four syllables. Because of this, I propose to
modify Chong's existing model to fill in these gaps.
4.1 Prefixes as Unique Prosodic Units
There are two modifications Chong's Model needs in order to accurately represent the data.
The first is that prefixes in CSE are treated as their own AP phrase. Others have discussed
how, in many languages, prefixes "fail to incorporate into the prosodic word to which they
attach" (Peperkamp, 1999). In other words, that prefixes form their own unique prosodic
unit from the stem that hosts them, morphologically. In the case of CSE, I argue that
prefixes are their own accentual phrase. This simple addition would account for words like
un-imaginable and bio-degradable.
29
150
# 100 e;
~ 75
aL I
0.2181
I I
HaL I
I I
HaaL HaL I I I
un
say the word
rna gl na
unimaginable
Time (s)
I I
HaaL I I
ble L H
agam
2.02691005
I II
HlL)i I II
2.04
Figure 15: Say the word unimaginable again spoken by consultant RT
2.61649946 400
• • • 350
300
~ 250 g il 0: 200
• ~
~ v'l
• i~
• If i* I I i~ l
. , • r ! .. " • : !-
I\! ~ ,,," , , • : ~ . -: • i-
., !"i • , • I I I I I I II
L H:nL HmL HaaL HaIL% I I I I I II
bi 0 de gra da ble
say biodegradable agam
0.7393 2.62 Time (s)
Figure 16: Say biodegradable again spoken by consultant PK
In Figure 15 there is a clear rise on the syllable un followed by an immediate fall as the
rest of the word begins. Chong's L* would not be able to account for the rising pattern on
30
the prefix. It is not just a low pitch accent but the restarting of a new accentual phrase.
Furthermore, there seems to be a perceptible juncture between the two syllables in question.
We see a similar process in Figure 16 spoken by consultant PK. Here, there is a clear rising
tone on the prefix bio with a sudden dramatic fall on the next syllable. It is more salient
to mark this change by using two separate APs rather than a low pitch accent. Thus,
I appeal to the idea that prefixes are prosodically separate, and propose that prefixes in
the intonational system of CSE should be treated as their own accentual phrase. This
modification is compatible with Chong's Model, as it only involves changing the definition
of what constitutes an accentual phrase.
4.2 Multisyllabic Words and the Accentual Phrase
In addition to the tendency for prefixes to be unique prosodic words, we must propose another
modification in order to account for words greater than four syllables. I propose that these
longer multisyllabic words which behave differently actually contain two accentual phrases,
as the intonation pattern of low to high seems to repeat itself. Whether or not these are
indeed APs and not some other smaller prosodic unit is unknown. However, it does seem
to be necessary to break down these multisyllabic words further in some way in order to
account for the pronunciation we are hearing, both in terms of intonation and juncture. In
both six syllable and five syllable words, the AP break appears in the middle of the word.
Again, this would not change Chong's model drastically, but rather modify how APs are
defined within it.
For example, we see this in the six syllable words memorabilia and developmentally. In
Figure 17 we see a rise in the first half of the word, followed by a break in the speaker as
31
the AP seems to restart the intonation pattern. This is heard in Figure 18 as well. When
listening to the recording the repeating rising pattern is clearly apparent. In both cases, there
seems to be a slight perceivable break in between the proposed APs in the pronunciation,
giving further evidence for the existence of a boundary within words.
209491007 350
",.. i. :~
'I 300
• 250
N ::s "" 0
• ~ \. ... ~~~ , ! •
~ 200 I I I
aL HaIL Ha aL HaIL HU' I I I I I I I
me mo ra bi Ii a
say memorabilia agam
0.3354 2.116 Time (s)
Figure 17: Say memorabilia again spoken by consultant PK
32
2.13451406 400
350 I' • • .. 300 •
N 250 e;
, J~
• \ ", ~ •
• • • "" . .@
200 p,
I •
I
~ HaaL HaaL HaaL Ha LL , I I I I I I I
de ve lop men tal Iy
say developmentally agam
0.2439 2.135 Time (s)
Figure 18: Say developmentally again spoken by consultant PK
In the five syllable word livability we have a similar problem with the prediction. Here,
it seems the second syllable of the word remains relatively low. We can not use a L* because
it is not the stressed syllable of the word. When listening to the recording we can hear a low
to high pattern occurring on the initial two syllables, and again on the latter three. In this
case, the first AP surrounds the first two syllables of the word li-va. The last three syllables
make up the next AP bi-li-ty. We can see this in Figure 19. It is important to note this is
the only example of this pattern in five syllable words within the available dataset.
33
1.35245227 200
150 - ~\ ~ .",,~ :.t ~
'fIOA.~~. ~ "1'.:
- .... " "
~! .. .. :- :
g 100
I I I I I I I I ~ p, 75
iL HaaL H:mL Ha aL Hab%
I I I I I I I I
Ii va bi Ii ty
say livability agam
0.04744 1.352 Time (s)
Figure 19: Say the word livability again spoken by consultant RT
4.3 Ternarity
Within the data set there seems to be a preference towards ternarity in accentual phrases of
esE. A ternary system is one in which prosodic units are grouped in threes, in contrast to
a binary system which groups prosodic units in two. Researchers have found more evidence
for binarity in prosodic structure than ternarity. Most syllable systems use binary feet,
composed of two syllables. (Goedemans, 2013). The fact that eSE may have a preference
for ternarity in building APs out of syllables requires further examination and discussion.
eSE's AP length preference is recorded in Table 3. The first column lists the number of
syllables in the target word. The corresponding column shows possible AP configurations,
separated by backslashes ( / ) 10.
10 Note that these are for attested words without prefixes.
34
There is a tendency in multisyllabic words greater than four syllables to break into two
APs, as discussed in Section 4.2. Within this tendency a pattern preferring three syllable
units becomes apparent. In six syllable words we see cases in which the word splits into
two three syllable accentual phrases. In five syllable words we see a similar split occurring,
manifesting itself as one two syllable AP followed by a three syllable one. Target words with
less syllables are never split into multiple accentual phrases. In a binary system, one would
expect a different outcome. In six syllable words, we would see a prefence toward three two
syllable APs. In five syllable words, we would have two two syllable APs and a lone single
syllable AP. Finally, we would also see two two syllable APs in four syllable words. In CSE,
four syllable words never seem to have multiple APs within them.
Not only do we see a preference for ternarity in this data, but also a preference for right
alignment. Five syllable target words within the data set that have two APs always have
the three syllable AP to the right of the word. It would be interesting to investigate if and
how right alignment manifests itself in other parts of Colloq uial Singaporean English. The
data set within this paper is very limited and only has two consultants. It is important to
investigate this matter further and see if it is a consistent pattern in more speakers alongside
more data. I will not be looking into it further other than to comment on it's tentative
existence here.
Another key fact from Table 3 is found in target words of four syllable length. Within the
data collected, these multisyllabic words are never split into two APs. One would predict the
possibility of a one syllable AP followed by a three syllable unit for these words. However,
this seems to be impossible. This may hint that CSE dis prefers one syllable AP, and only
uses them when it is necessary-for one syllable words and prefixes. This could also lead into
35
an investigation on the ordering for these AP rules. An examination of this data through
the lens of Optimality Theory would be an interesting topic for a future paper.
# of Syllables in Target Word
6 5 4 3 2 1
# of Syllables in AP(s)
( (5(5(5 ) ( (5(5(5 ) / ((5 (5 (5(5 (5 (5 ) ((5(5) ((5(5(5) / ((5(5(5(5(5) ((5(5(5(5) / *((5)((5(5(5)
( (5(5(5) ((5(5 ) ((5)
Table 3: Target Words and Corresponding Syllable AP Length
4.4 Exceptions
There are, however, exceptions to these modifications. For example, there are certain cases
in which the target word developmentally does not exhibit this two AP pattern. In Figure
20 we see consultant RT's pronunciation of the target word does not have two APs, nor
does it have a low pitch accent. Instead, it has a consistent rising action throughout the
entire word. One possible explanation is rate of speech. It could also be due to the fact the
speaker seemed a bit distracted as he pronounced the data, hinted at with the unusual pause
between the target word and the word again. This could also be due to speaker variation.
Lastly, it could just be an exception to the rule. However, it is important to acknowledge it.
As it stands now, these two modifications surrounding prefixes and multisyllabic words
help strengthen Chong's current model and better represent the intonational system of Col-
loquial Singaporean English as we know it. His model remains relatively in tact, with only
a call for further investigation on what exactly constitutes an accentual phase. These modi-
fications are in need of more data to be fully substantive. Moreover, more data is necessary
36
2.05195317 2001~--~~~--~--~~----~----~~------~--~--~~
\ •••
· ·w " .. ~ ~ .~,. 4: "\ ""';.r, '. , ,.... n.,:. • • . \ :. : : :
•
"
150
100
75n---T-II~--~~--~--~----rill~----~,~----r1
PL HaaL HL- aL Ha
II I I I I I
de ve lop men tal ly
say developmentally [pause] agam
0.201 2.052 Time (s)
Figure 20: Say developmentally again spoken by consultant RT
to explore CSE's apparent tendency for ternary systems.
5 Conclusion
In Colloquial Singaporean English there is a general intonational rise that occurs over the
course of a single prosodic unit, usually the size of a content word. This is found with
interspeaker consistency. This finding confirms both Chong and Ng's Model, as each predict
this rising pattern. One gap in the literature, however, is how multisyllabic words behave
within this system. In this thesis, I com pared both models and their ability to predict longer
multisyllabic words accurately in order to improve upon this subject.
From the data collected it is apparent that Chong's Model is currently the most accurate
at representing the dataset. Ng's Model is inadequate when describing single syllable words
37
and multisyllabic words with a low pitch accent on the stressed syllable. Chong's Model
remained consistently able to predict multisyllabic words of longer length, but is in need of
minor modification in relation to what an accentual phrase is consistent of.
In Section 4, I proposed two modifications to Chong's Model in order to better describe
the data set. This included a modification that treated prefixes as their own prosodic unit
separate from the word they are attached too. In addition, I proposed that larger words,
such as six syllable words, may actually contain two APs rather than one. However, in order
to substantiate these claims more data is necessary.
One avenue to take in investigating the nature of APs in CSE is to look at boundaries
within words such as cannot. Cannot is treated as its own word in Ng's analysis, as stated
in Section 2.3. It would be interesting to look into this further and see if other similar words
behaved in this way, such as will not. This may shed more light into the nature of AP
boundaries.
An important aspect of this research to note that has not been brought up is the behavior
of the word say. There are times in which it is in the same AP as the following word, as
there is no perceptible rise on it, pointed out to me by Chong. This is problematic, as it is
its own content word and should be have consistently throughout the data. This behavior
could be explained by the monotonous task of the elicitation, as each sentence begins with
the word say every time. It could also hint again towards CSE's dis preference toward single
syllable APs. It is an aspect of this research that deserves further probing.
During this investigation, a tendency towards ternarity made itself apparent in CSE's
intonation system. In multisyllabic words with two APs, the split always occurred so that
the final three syllables remained a unit. This preference for three syllable units seems
38
typologically unusual and deserves its own further exploration. Does this preference exist
in other instances of CSE? If so, where? Is this another borrowing from a language it is in
contact with, or did it come about by itself? In addition, there seems to be a preference
toward right alignment as well. Is this apparent in other aspects of the language? These
questions are outside the scope of this paper but would make excellent topics for further
investigation.
Furthermore, another important question is still left unanswered. What type of lan
guage is Colloquial Singaporean English? It does not behave as a lexical tone language like
Mandarin Chinese, as it does not have lexical minimal pairs differentiated only by tone. In
addition, although stress is important to it's intonation system (Ng uses it to ascribe tone to
a syllable, Chong uses it in his definition for low pitch accents), it does not seem to have any
prominence lending abilities like stress-accent languages have. This leaves CSE as a pitch
accent language, similar to Japanese. The low pitch accent in Chong's model does seem to
hint at this, as it is ascribed to the lexical word based on its stress pattern. However, there
is too much variability within the speakers to see this as a consistent aspect of the language.
This leads me to believe that Colloquial Singaporean English exists outsides of these three
options. This could be because it has been influenced by so many languages of different
intonational type. More data is needed to investigate this matter further.
It is important to revisit the conclusions of this paper with more data, specifically data
of longer multisyllabic words. In my initial investigation, I elicited a large array of target
words with varying syllable length. In the future I would look more closely at just longer
multisyllabic words to search for more patterns and see if there is a bias towards ternarity. In
addition, it would be important to consult with more speakers. It would also be interesting to
39
look at other kinds of morpheme boundaries and how they play into the intonation systems.
Finally, another avenue of thought would be to examine an OT analysis of accentual phrase
boundary ranking. Especially to see how much one syllable APs are dispreference. One
could test this with short words with prefixes and so forth.
In searching for answers to multisyllabic words in eSE's intonation system we have dis
covered an array of more questions, from ternarity to AP boundaries. All of these questions
are deserving of answers, but are unfortunately outside the scope of this paper. More data
is necessary to continue the search for these answers.
40
6 Appendix
350
300
250 'N' e;
"' ~ 200
aL I
0.04141
210
150
g 100
il 0: 75
aL I
0.04197
0.930174478
v • • • • ••
.- .. HaaL Ha aL HaHH%! I I I I I I I
H H L H
say Min agam
0.9408 Time (s)
Sentence (1) spoken by PK
1.06643871
• , I' f •
~V:N ,
J I I I I I I
Ha aL Ha aL HaHH% I I I I I I I
H H L H
Say Min agam
1.075 Time (s)
Sentence (1) spoken by RT
41
0.991096253 410
./ , ; , 350
300
I' g 250
"' • 200 I
aL HmL Ha aL HaH-% I I I I
I "
H ,H L H
say live agam
0.4174 1.632 Time (s)
Sentence (2) spoken by PK
1.04292816 170
160 ~ •
140 /""" ... ~
'-120
.. " • • -N '~ .~ ....... "' '\.,,; • J
'N' 100 e;
"' ~ 75 I I I I
aL Ha aL Ha aL Ha LL% I I I I I I I I
L H L H
say live agam
0.324 1.043 Time (s)
Sentence (2) spoken by RT
42
1.77486495 400
350
~ 250 g /: •
\ .. '
~: ;.~\ . ..
300
• • , , , , 200
aL HaaL Ha aL HaaL Ha...b°: , , , , , , ,
, , "
H L H H L H
say the word bee agam
0.3268 1.781 Time (s)
Sentence (3) spoken by PK
1.56445021 170
160 , 140 • 120
'N' 100 e;
~:--V\ -'~ I ...A • . ~ ,
"' ~ 75 , , , , , " , .
L HaaL HaaL HaaL I-lliL%i I I I I I II I
H L H H L H
say the word bee agam
0.4637 1.571 Time (s)
Sentence (3) spoken by RT
43
2.05743383 0.676029873 450~---------,~--,------,,,--------,--,,-------~
400
350
300
g 250
~ 200
aL I
H
say
0.676
170 160
140
120
# 100 e;
"' ~ 75
L II
H
say
0.443
H:nL I
L H
the word
I
Ha aL I I
Time (s)
I I I
Ha L-L% I I I
H
may
Sentence (4) spoken by PK
2.057
1.22704389
~ '''~V: .: ! '-", Ioi
I I I I I
HaaL HaaL HL-L%! I I I I I I
L H H
the word may
1.233 Time (s)
Sentence (4) spoken by RT
44
0.313968195 400
350
300
~ 250 g 200
aL I
0.314
200
; .. .. 150
.. 100
75 I
rL I
0.04692
",-,"': • ••
\ ... ." \
.~!~ :- •
I
HaaL HaaL I I I I
H M H L
say rrnm
Time (s)
Sentence (5) spoken by PK
, . .,~A~ ..
\ • ,...-v , iUr :* • .: ;
I I I I
HaaL Ha aL I I I
H M H L
say rrnm
Time (s)
Sentence (5) spoken by RT
45
Ha LL% I I I
H
agam
1.902
0906330083
• • ","
~
I
HalLl
I "
H
agam
0.9063
1.34018958 400
./ 1 ,.,. 350
300
~ 250 g """ , 200 I
aL Ha aL Ha aL Ha LL% I I I I I I I I
H M H L H
say living agam
0.5104 1.842 Time (s)
Sentence (6) spoken by PK
1.07977607 170
160
.. .
140
120
'N' 100
~ ~ """-"'\ .-::~, .. . • ,. ~ .
~
. ~
e;
"' ~ 75 •
I I I I I II
L HaaL Ha aL Ha lL% I I I I I II
H M H L H
say living agam
0.2252 1.08 Time (s)
Sentence (6) spoken by RT
46
# e;
~
2.1190729 350~--------CT--'-----CT-------'-------'-''-'-----'-~
300
250
200
L
H
say
0.612
190
150
I • :. ~
100 •
75 I
aL I
H
say
0.09771
I I
HmL I
L H
the word
I
HmL I
~ " ' , :
L
I ,
believe
Time (s)
,
H
t
Ha aL I I
L
I,
Sentence (7) spoken by PK
" • • • •
"'" •
,,/' • • • " , ,
.. ~. " ~' I .. . ! .,,';""t: .'\ :: : : '\
I I I I
HaaL Ha aL HaaL I I I I I
L H L H L
the word believe
Time (s)
Sentence (7) spoken by RT
47
•
H
HaL I I
agam
2.133
1.35244717
' . ti,
I
Hhb% I I I
H
agam
.382
1.83195755 410
350
g 250 /[1\ '" 300
200 I I
aL HmL Ha aL HaL-L%! I I I I I I I I
H L H M H
say the word maybe
0.6639 .849 Time (s)
Sentence (8) spoken by PK
0.955681915 170
160
I 140
120
'N' 100
, .,w~
I
~ .... 4'10.1
e;
"' ~ 75 I I I I
aL HaaL Ha aL Hib% I I I I I I I
H L H M H
say the word maybe
0.1585 0.9557 Time (s)
Sentence (8) spoken by RT
48
2.46925779 400
350 A .\
" ,
I • • 300
~ 250 g
" 200 I I
fL HaaL HaaL aHaL HaIL%! I I I I I I II
H L H M H L H
say the word lInage agam
0.8333 .522 Time (s)
Sentence (9) spoken by PK
1.2456212 170
160
• 140 ~-.
•
120
'N' 100 " \
N.~ \ • • ,
:\ i\ • ;.,"'"
• "" , : -i·
e;
"' .. ~
"" ~ 75 I I I I
aL Ha aL Ha aL Ha aL Hh-L°o I I I I I I "
H L H M H L H
say the word lInage agam
0.1354 1.246 Time (s)
Sentence (9) spoken by RT
49
1.86918161 400
,.: $ ~ •
1 il
350
300
\11 • ~ 250 g ,. , ..
I 200
aL HaaL Ha aL HaLb% I I I I I I I
H M M H L H
say minimal agam
0.3856 .88 Time (s)
Sentence (10) spoken by PK
1.42683292 170
160
140
120
'N' 100 e;
~y~W'''''--'~) • , • , .. ' : . : ,:: , i·" , .,
"' ~ 75 I I I I I I I I
aL HaaL Ha- aL HlLL% I I I I I I I I I
H M M H L H
say minimal agam
0.4957 .461 Time (s)
Sentence (10) spoken by RT
50
400
350
300
~ 250 g 200
aL I
0.4137
170
160
140
120
100
75
L II
0.1891
~~j "', .. ~
H
say
I I
HalL I I
H
say
'- • !, ~
I I I I
HaaL L* Ha aL I I I I
M M H L
livable
Time (s)
Sentence (11) spoken by PK
M M H
livable
Time (s)
I I I
HaL- aL
I I
L
Sentence (11) spoken by RT
51
"-
H
agam
x
agam
1.77035762
HaL% I II
1.77
1.19901924
X I
1.199
2.39144676 350
A' ~ 'l A \~ • :'1 • --.;~ • •
300
250
• i
200 I I I I I
L HaaL HaaL L* Ha aL Ha LI, , I I I I I I I I I
H L H L M H L H
say the word believer agam
0.8014 .418 Time (s)
Sentence (12) spoken by PK
1.36897779 170
160
140
120
'N' 100 e;
,
.~ N~ ~ 1M !~
, \ :: . • • , i' i~
"' ~ 75 I I I I I I I
rL HaaL HaaL L* Ha aL H:lL% I I I I I I I II
H L H L M H L H
say the word believer agam
0.09495 .392 Time (s)
Sentence (12) spoken by RT
52
2.01876843 410
:, '" .; \. i~
350
300
g 250 '-' ..... 'i
200 I I I I
aL HaaL HmL L* Ha-I-H% I I I I I I I
H L H M M H
say the word rnaybelline
0.6929 2.019 Time (s)
Sentence (13) spoken by PK
1.05371212 170
160 • :.
140 • ,.
l • 120
'N' 100
! ;.,--,.
.. '"""-~ ),IV . N~ . .J. : :
•• " e;
"' ~ 75 I I I I I I I
L HaaL Ha aL L* HalL,
I I I I I I I
H L H M M H
say the word rnaybelline
0.2061 1.054 Time (s)
Sentence (13) spoken by RT
53
400
350
300
~ 250 g 200
L
0.2917
170
160
140
120
# 100 '. e; • "' ~ 75
,~
I' L
II
0.2059
;l \ ,
H"L ,
H L
say the
•
H
say
, ,
Ha aL I
L
the
\ t
f\ ,-.,
,
HalL HL-aL , , , , ,
H L M H
word nnagme
Time (s)
Sentence (14) spoken by PK
•
H
, ,
HaaL I
word
L M
nnagme
Time (s)
H
,
HalL I I
L
Sentence (14) spoken by RT
54
L
2.09748404
Ha Ib%
, "
H
agam
2.097
1.49820503
,
Ha LL% I I I
H
agam
1.504
1.83542132 400
350
~ 250 g
• - .....,ir • -.. -
,~r : . •
,,- , . , . • •
300
200 , , , , ,
aL HaaL L* HaaL Ha LL'i< , , , , , ,
, "
H L M M H L H
say minimally agam
0.543 .864 Time (s)
Sentence (15) spoken by PK
1.55780704 170
160
140
120
!~~ /. ,
'N ' , :" ! ! •• '" • : :
'N' 100 e; "" l
"' ~ 75
.,. , , , ,
L HmL Ha aL Ha LL% I I I I I
H L M M H L H
say minimally agam
0.3549 1.571 Time (s)
Sentence (15) spoken by RT
55
1.32766671 350
.,l \" ,/' I ,
~ \
"... .:
300
250
•
200 , ,
aL Ha aL HaaL Ha aL Ha lb~,
, , , , , , , , ,
H L H L M M H L H
say the word believable agam
0.6235 2.383 Time (s)
Sentence (16) spoken by PK
1.71569971 170 160
140 • •
120
'N' 100 " . r. .
"'~~ .. ~~ M , ' , " I • • e;
"' ~ 75 , , , ,
'i' L H:nL HaaL HaaL HUIo
I, , , , , , , ,i,
H L H L M M H L H
say the word believable agam
0.2495 .756 Time (s)
Sentence (16) spoken by RT
56
0.742595595 400
350 Ii: I :' n • I ,
• •
300
~ 250 g • , ... , " · , " 200 I I I
L HaaL Ha aL HaaL H:lb , II I I I I I I I
H L H L M M H L H
say the word nnagImng agam
0.2249 1.959 Time (s)
Sentence (17) spoken by PK
1.47510762 190
" • ,
g 100
• : : • : . : : • : -: : . : :
i \! ~ •• r:'4~ : "'~ ! M"")). "("" ., ~ , "I, , . .
·"IH . .
150
75 I I I I I I I I I
L HaaL HaaL HaaL Ha LL% I I I I I I I I I
H L H L M M H L H
say the word nnagImng agam
0.197 1.475 Time (s)
Sentence (17) spoken by RT
57
0.783858091 400
350 A ... I
• 300 \ ,.... , ~ 250 g "' .. ...,.
200 I
aL HaaL HaaL HaL-L 0
I I I I I I
H L M M H L H
say malaria agam
0.3115 1.585 Time (s)
Sentence (18) spoken by PK
1.29246636 170
160
140 "\ • .. • -.. ii, 120 'ri.I~."..J- u N ii ; ,
'N' 100 e;
"' ~ 75 I I I I
L Ha aL HaaL !lib');
II I I I " H L M M H L H
say malaria agam
0.2511 1.292 Time (s)
Sentence (18) spoken by RT
58
0.485855276 400
350
300
~ 250 g 200
fL
0.4859
170
160
140
.I ,
HaaL , ,
H
say
" ""- I'''
Ha ,
L M M H
rornama
Time (s)
Sentence (19) spoken by PK
i "1;\ 120 ~-~ : - .
i" • 4 -. :-
r i ' :
'N' 100 e;
"' ~ 75 , , , , ,
aL Ha aL HaL- aL , , , , ,
H L M M H
say Romania
0.0709 Time (s)
Sentence (19) spoken by RT
59
aL Ha L-L 0
, , ,
L H
agam
2.196
1.15627799
: ~:
,,~
:.:
~~
" Ii
"' HIL%
"' L H
agam
1.156
2.28268096 400
350
300 ~\i
~ : r ~ 250 g ", ~ ,
• 200 I
aL Ha aL L* Ha aL Hm% I I I I I I II I
H L M M M H L H
say livability agam
0.4858 2.284 Time (s)
Sentence (20) spoken by PK
1.34182703 200
: :
: ~\
~ -~."...",......., -"'" . .: : \1':,1 .. :- :
150
100
75 I I I I I I
L HaaL Ha aL Hdb% I I I I I I I
H L M M M H L H
say livability agam
0.04744 1.352 Time (s)
Sentence (20) spoken by RT
60
2.20982006 400
350
/1' "'\ \ : .: I
300
V ! !i ~ 250 g 200 I I
'" HmL
I I
H L
say the
0.07826
200
150
""""'" 100 ~" .
•
~-I I
L HmL L* I I I
H L M M M M
word unimaginable
Time (s)
Sentence (21) spoken by PK
• ,
.. I I
HaL- aL HILb 00
I I I I I
H L H
agam
2.21
2.05286242
75n---~,,~~-r~~--~~I~--~~--~I~~I------rll~,~
L II
H
HaL HaaL L* HaaL HJi,%
I I I I I " I
L H L M M M M H L H
say the word unimaginable agam
0.2287 2.053 Time (s)
Sentence (21) spoken by RT
61
400
350
300
~ 250 g
200
L II
0.2439
170
160
140
120
100
75
L II
0.201
H
say
I' .. • • • c \
I
HaaL I
H L
say
" I HaaL
I
L M
1.68349262 2.13133129
•
J'1 •
", ~ f
• , , I
I
L* ffiaaL Ha LL , I I I I
M M M M H L H
developmentally agam
2.131 Time (s)
Sentence (22) spoken by PK
2.03717583
, v ~~~ '.' "
: : .. .
I I I
HL- aL Ha I I I I
M M M H L H
developmentally [pause] agam
2.041 Time (s)
Sentence (22) spoken by RT
62
350
1"1'
• /!~ I • i- i •
300
250
200 I I
aL HaaL I I
H L
say the
0.4882
0.363532471 170
160
140
"\ • , : ,-,
"-" '-.' ) • II! • I I
HaaL L* I I I
H L M M M
word developmentally
Time (s)
Sentence (23) spoken by PK
, '" · .. ~ ! -y\"" !
l"'~ , ,(, .......... '" ;
i# i·" iii ~ 120
• 100 \ •
• 75 I I I I I
L HaL Ha aL L*
I I I I
H L H L M M M
say the word developmentally
0.3635 Time (s)
Sentence (23) spoken by RT
63
2.23048629
•
• I
• • ,
HalLOo I I
M H
2.23
:"'\j. ., (['
!'.~ i#· ,
• • •
• ! .-I
Hil.-~
II
M H
1.619
350
300
250
200
aL
0.332
0.09390093 170
160
140
\
,;- i~ :.,j
I •
I
HmL I I
H
say
,....,/A i1
120
100 J
\
75 I I
L HaaL I I
H L
say
0.0939
2.10206726
•
\.. ~ ...;;: -.. .. , '" . ,
~ f
I I
L* HmL HL L , I I I I
L M M M M H L H
memorabilia agam
2.102 Time (s)
Sentence (24) spoken by PK
• " M-"'-~-\~~ , , , ,
: : ::: ,
'\Ii
I I I I
L* HmL Halb°o I I I I
M M M M H L H
memorabilia agam
1.363 Time (s)
Sentence (24) spoken by RT
64
0.455539144 400
350 ...
~ 250 g
A' ::~ ~ • .1 •
~ ~ •
It
300
'. 200 I I I I
L Ha aL Ha aL I I I I
H L M M M H L
say radiology
0.4555 Time (s)
Sentence (25) spoken by PK
170
160
140
120
::: :
':~ • . .
~ • h, : \ ~ : :! •
'N' 100 e;
• • •
"' ~ 75 I I I I I I
L HaaL Ha aL H:nL I I
H L L M M M H L
say ra- [pause radiology
0.3271 Time (s)
Sentence (25) spoken by RT
65
2 13550039
: I •
• • : '.:
:'-'l~ ." I I
Halo , I I
H
agam
2.136
2.01804023
"'"'" " .11 I
, i
I II
HiIik% I II
H
agam
2.018
300
250 'N' e; ~ ii: 200
say
0.7628
• 140
120
,
ii,
~ {H ~I ; Ha aL
H L
Time (s)
b PK (26) spoken y Sentence
2.60972023
., : ii it ~: ~ :. . .
~a ~% i
H
2.7023636
'1lrfiii~ m·1t~ :: :-
g 1001!, ____ -rr-____________ ~--------~~~~~~~~-% ~ ~ "" .
~* Ha aL
75
HaaL
-.l-.l
aL
-.l -.l-.l -.l
H L
say
1.318
M M M M H
Time (s)
b RT (26) spoken y Sentence
66
L H
REFERENCES
Boersma, Paul & David Weenink. 2015. Praat: doing phonetics by computer [Computer
program]. Available at http://www.praat.org/
Chong, Adam. 2012. A preliminary model of Singaporean English intonational phonol
ogy. In Working Papers in Phonetics No. 111, p.41-62.
Chong, Adam. 2013. Towards a model of Singaporean English intonational phonology.
In The fournal of the Acoustic Society of America Vol. 133 No.5, p.3570.
Chong, Adam & James Sneed German. 2015. Prosodic phrasing and FO in Singapore
English.
Gunlogson, Christine. 2008. A question of commitment. In Belgian fournal of Linguis
tics, 101 7136.
Harada, Shinichi. 2009. The roles of singapore standard english and singlish. In Infor
mation Research 40, 70-82.
Ladd, Dwight Robert. 2008. Intonational Phonology. 2nd ed. Cambridge University
Press.
Leimgruber, Jakob. 2011. Singapore English. In Language and Linguistics Compass 5.1,
47-62. Blackwell Publishing.
Ng, E-Ching. 2011. Reconciling stress and tone in Singaporean English. In Asian
Englishes: Changing perspectives in a Globalized World.
Ng, E-ching. 2012. Chinese meets Malay meets English: origins of Singaporean English
word-final high tone. In International fournal of Bilingualism.
Peperkamp, Sharon. 1999. Prosodic Words. In GLOT International 4.4, 15-16.
Goedemans, Rob & Harry van der Hulst. 2013. Rhythm Types. In The World Atlas of
Language Structures Online.
Tay, M. W. J. 1993. The English language in Singapore: Issues and development. Sin
gapore: U niPress.
Venditti, Jennifer J. 2005. The LToBI model of Japanese intonation, in Sun-Ah Jun
(ed.), Prosodic Typology: The Phonology of Intonation and Phrasing. Oxford:
Oxford University Press, 172-200.
67